Charge coupled device (CCD) cameras are commonly used in digital imaging systems. Multi-tap CCD cameras split an image frame into two or more areas that are read out from the CDD camera in parallel. For example, an image frame of a multi-tap CCD camera may be divided into left and right halves, and the pixels in each half are read out from each half in parallel electrical paths. These multi-tap CCD cameras exhibit inherent response errors between the readouts of the different paths. For example, PVT (process, voltage, and/or temperature) variations in each path may produce errors. As a result, there may be differences in the gain and/or black levels of each group of individually read out pixels from each path.
As a result of poor tap/channel matching between paths of the multi-tap CCD camera, the difference in the pixels read out from each path may result in a “split screen” visual effect when the pixels from each path are combined into a single image. In an image exhibiting a “split screen” visual effect, a relatively clean line of demarcation exists in the image between pixels read out from each half of the image frame, as a result of the differences in the pixels read out from each path. Digital image processing also typically performs contrast-enhancing and/or peaking operations on the image frame which may increase the error(s) between pixels from each path.
Conventionally, errors between parallel readouts in multi-tap CCD cameras are addressed though single-point or multi-point factory calibration performed by the camera vendor. For example, one approach has been to attempt to minimize tap-based errors by performing a simple level calibration, assuming a generally-well balanced camera as an initial condition.
Some factory calibration approaches manipulate digital controls of the analog to digital converters associated with the pixels. For example, U.S. Pat. No. 7,236,199 (hereinafter the '199 patent) discloses performing automated tap balance within a CCD camera by controlling the analog to digital converter dynamically. In this method, gain resolution is limited by the control resolution and PVT variance of the analog to digital converter, and as a result, the minimum gain error is at least several digital counts. In addition, the method disclosed in the '199 patent requires the CCD sensor to be shrouded to perform level calibration.